[Gmsh] meshing differences between version 1.65.0 and 2.0.4

David Colignon David.Colignon at ulg.ac.be
Wed Mar 21 12:47:33 CET 2007

Daniel Wheeler wrote:
> Thanks for the prompt reply.
> On Mar 20, 2007, at 5:09 PM, David Colignon wrote:
>> Hi Daniel,
>> yes, there are some differences in the default treatment of the 
>> characteristic lengths ( lc ) between version 1.65 and 2.04.
>> In version 2 we added the ability to import iges or step files and in 
>> this case there is no characteristic length associated to the points, 
>> so we have to choose an arbitrary default value. The choice is 
>> difficult because we have to be able to mesh very small as well as 
>> very big objects, and also be able to merge step files (no lc) with 
>> .geo files (imposed lc). And there is also the problem of imposed 
>> background meshes and of attractor points.
>> We are still looking for he best compromise ...
> Fair enough.
>> We end up with a value for lc that can not be greater than 1/10 of the 
>> largest dimension of the bounding box (to be able to mesh step files).
> I am trying to create a tapered mesh from a fine to a coarse region. If 
> I want one edge to have only one or two elements then the
> rectangle must have a length that is 10 times greater than the width. Is 
> that correct?

Yes, but I suppose it is not a good solution for you

You can also ad an isolated point 

Point(999) = { 10 , 10 , 10 , spacing } ; 

to artificially increase the size of the bounding box and get rid of this limitation.

>> If you want the same kind of behavior in version 2.04 as in version 
>> 1.65, in _your_ case, you have to modify line 194 in 
>> Mesh/BackgroundMesh.cpp : 
>>  double l3 = CTX.lc / 10.;
>> becomes
>>  double l3 = CTX.lc / 1.;
>> and recompile.
> Not an option. The code is distributed to others. Could the divisor be 
> made a parameter that is passed?

It could be... We are still looking for better solutions to this problem